Add support for testing the WebExtensions API in WPT

[kiaraarose]

No description provided.

[gsnedders]

We should probably care to some extent about how easy/difficult it is to write tests without using extension.js.

Rather than putting all the code necessary to map from browser.test to testharness.js in the server-generated wrapper, it would probably be better to put it into a separate file — which then gets included (along with testharness and testdriver) by the server-generated wrapper?

[OrKoN]

I have actually tried to implement the proposal for executorsafari.py in web-platform-tests/wpt#50648 (comment) and I think it would be easier to integrate the tests without adding extension.js and using regular testharness.js tests. I think something like this would work:

// META: script=/resources/testdriver.js?feature=bidi
// META: script=/resources/testdriver-vendor.js
// META: script=/web-extension/resources/webextension-test.js

webextension_test(async () => {
    // Call for illustrative purposes, in practice, webextension_test
    // defined in webextension-test.js can implement it. 
    await test_driver.bidi.webExtension.install();
}, test_desc);

A problem with the current approach is that the infrastructure does not detect the ?feature=bidi inside the server-generated wrapper.

[OrKoN]

here is a runnable implementation of this web-platform-tests/wpt#51090

[jgraham]

Can we get an update here to detail the current proposal? At the moment what's in the RFC doesn't seem to match what's actually in the patch.

My understanding is that the proposal is for a new test wrapper indicated by .extension.js in the filename, a new testdriver API, and a function exposed by the wrapper that implements a protocol for transferring test results back from the extension? In particular the latter protocol needs to be specified somewhere (I assume it's what browser.test does, but unless that has a written specification somewhere we need enough detail in the RFC that one can understand what the requirements are without reading Chromium-specific documents).

[OrKoN]

I have concerns that we will end up with two methods for dealing with extensions exposed to test authors following different (and separately evolving) specs that are not fully implemented by all vendors. I think it would be preferable to start with adding methods to testdriver.js that match the WebDriver BiDi specification and allow synchronous handling of these methods for non-WebDriver-BiDi implementations (prototype: web-platform-tests/wpt#51090).